I have a design problem. That's to design a wavelength locking loop for a DWDM optical transmitter. The DFB laser source used in the transmitter has a wavelength temperature coefficient of .106 nm / degree centigrade. The temperature of the laser chip cant be allowed to drift more than 0.1 degree centigrade.

I'm using two feedback loops for achieving wavelength control. One is a thermal feedback loop with a thermistor and thermoelectric cooler which are within the laser chip. Another is with a wavelength sensor which is afabry perot etalon based device. Which has two photo detectors one of which recieves the optical input directly and the other through an etalon. The difference in the photodetector currents gives a measure of the drift in wavelength.

I intend to use PID control for both loops(Thermal as well as wavelength).Could someone suggest some references where i could get some understanding about how to model the plants(Thermal as well as wavelegth sensor) and obtain their transfer functions so that i can tune the PID parameters.

 

It sounds like you are about to make a standard cascaded control.

1. you measure wavelenght drift.
2. you feed this into one PID (PV)
3. the output from this PID is feed into the second as SP
4. the measured temperature is feed into the second as PV
5. the output from the second PID control the amount of cooling

Am I right so far about what you are trying to do?


First question:
Do you really need the temperature transmitter and the second PID?
To answer this you should study response time for both transmitters, you should really use both only if the temperature transmitter is faster than the wavelength transmitter. To check that you should change the cooling signal manually (ie
make a step response). I seriously doubt you will get an answer a lot faster on the temperature sensor.


Ok, assuming you do get a faster response on the temperature sensor: Connect controllers as described above, note that the second controller need to be direct acting.

Optimize tour temperature controller, depending on what optimization method you want to use you could use the previous step response.

Optimize your wavelength controller, if you use a step response optimization method you have to make a new step response counting in the tuning of the temperature controller. The tuning of this controller should be made slow, I would reccomend
you use a method where you can make the tuning slow such as lambda.


Ok, assuming you don't get a faster response on your temperature sensor. Skip the temperature controller and make the wavelength sensor
directly control the cooling.



/Johan Bengtsson

----------------------------------------
P&L, Innovation in training
Box 252, S-281 23 H{ssleholm SWEDEN
Tel: +46 451 49 460, Fax: +46 451 89 833
E-mail: [email protected]
Internet: http://www.pol.se/
----------------------------------------

 

Dear Johan,
Thanks for your valuable suggestions. All the same, I am sorry to say that my query was not properly phrased originally. I am not going for a cascaded control. What I meant by saying that I am using two feedback loops is that I am using only one of these loops at a time. Initially at startup, or whenever the laser is off, there is no optical signal, so the wavelength locker loop cannot operate. Also the capture range of the wave locker device is such, that it can only stabilise the wavelength, if the temperature is preset to within +/- .5 deg. C of the exact setpoint. Therefore the thermistor feedback loop will operate alone at power up, with a fixed setpoint. Once the temperature is within the above mentioned limit, the wavelocker loop takes over. At that time the thermistor loop is not operating. Whenever laser is switched off, again the thermistor loop takes over.
Now I have the following queries:

1)Can we use the same PID controller block for both these feedback loops, in which the error signal is switched from one source to the other or will it be advisable to use seperately optimised PID blocks for the 2 loops?

2)Can the switching be through an analog switch, or some slow switching mechanism is required?(Most probably we will go for PI controller which anyway has a slow step response!)

3)Are there some standard mathematical models of Thermo Electric Coolers(TEC)? Which would help us to determine the step response of the "Plant"? In other words, how do we get the relevant parameters for tuning the PI/PID block?

With best regards,
Sameer